1. Field of the Invention
The present invention relates to a system and method of validating equalizer training.
2. Description of the Related Art
Modems are communications devices, which employ digital modulation techniques to transmit binary data over analog band-limited communications channels. A source modem collects an integral number of bits of data at a time and encodes the bits into symbols for transmission at a signaling rate. In bandwidth-efficient digital communications systems, the effect of each symbol transmitted over a time-dispersive channel has extended beyond the time interval used to represent that symbol. The channel distortion caused by the resulting overlap of received symbols is termed intersymbol interference.
Intersymbol interference has typically been compensated through channel equalization by an equalizer of a receiving modem. The equalizer is traditionally a finite impulse response (FIR), or transversal, equalizer based on a time-varying impulse response of the communications channel. The equalizer is also commonly a least mean-square (LMS) equalizer configured to minimize mean square error (MSE)xe2x80x94the sum of squares of the intersymbol interference terms plus the noise power at the output of the equalizer.
Equalization has typically involved an equalizer training sequence. A training sequence or train is an initial known burst of a symbol stream transmitted over the communications channel. Based on the training sequence, a training algorithm ideally converges the coefficients (or tap gains) of the equalizer to optimally compensate for channel distortion. In some cases, external tone tampering or signal dropout during training will corrupt the training sequence and, thus, the algorithm will falsely converge the equalizer coefficients to bogus values. This is termed false equalizer training. A primary channel of a modem typically crashes as a result of false equalizer training.
Briefly, false training may be detected by techniques to validate equalizer training. The validation techniques involve checking the unique time domain energy shaping of a plurality of taps of a trained equalizer of a modem for good training. The energy measure for each equalizer tap of the plurality of taps is approximated by summing the magnitude of a real component and a magnitude of an imaginary component of an equalizer coefficient corresponding to the equalizer tap. A tap with a highest approximate energy (i.e., the main tap) is then determined. Next, the main tap energy is compared to a main tap threshold and a predetermined guarding threshold. The plurality of taps are arranged in a time index ordered sequence. In determining the main tap threshold, a first energy sum of a beginning set of taps in the sequence is compared with a second energy sum of a last set of taps in the sequence. The lesser tap energy sum between the first energy sum and the second energy sum is the main tap threshold. The main tap is then compared with the main tap threshold and the predetermined guarding threshold. If the main tap is greater than the main tap threshold and the predetermined guarding threshold, then the train of the equalizer is good. If the main tap is less than the main tap threshold or the predetermined guarding threshold, then the train of the equalizer is false. The above-mentioned validation techniques are effective to recognize the unique time domain energy shaping characteristics of good-trained equalizer taps.
An approach to implementing the validation techniques is through software executed by a data pump of the modem; an application of the validation techniques is to improve V.34 half-duplex Phase 3 equalizer training; and an advantage of the validation techniques is a reduction of a likelihood of channel crashes.